You might not think it, and some of my critics will be laughing into their cornflakes at this point, but I spend quite a bit of time thinking about prose. It's fair to say that a sizable proportion of science fiction readers don't much care, or at least don't think that they care. Read a sample of amazon reviews of science fiction novels - in fact, a more general sample of web reviews will do just as nicely - and you'll find a lot of stuff about the plot being interesting, the characters likable, the story fast paced, the world-building good and so on. Or, indeed, quite the opposite - they didn't care for the plot, they couldn't relate to the characters ("I didn't want to spend any time with these people"), the story dragged, the world-building was insufficiently thought-through, and so on. Quite often you'll see a statement to the effect that the book was "well written", but as often as not there's no deeper qualification as to what it meant by this. Usually, I would contend, it simply means that the pages slipped by effortlessly enough, that the story was adequately engaging, that there were some good bits and an ending that was both comprehensible and satisfying.
A reviewer of mainstream or literary fiction, though, probably has something slightly different on their mind. Whereas a genre reviewer might take "well written" to mean a quality of transparency - functional, efficient, prose that doesn't occlude the narrative - the mainstream reviewer is probably applying a somewhat different set of criteria. What they mean, generally, is that the prose aspires to be more than merely a painless delivery mechanism for the story; that it can and should do more than that. Not being clumsily written doesn't get you bonus points: that's the absolute least that should be expected. Nor is it enough to avoid cliches; that's only half the job. We can all omit cliches, find prosaic workarounds that convey the same sentiment - but that's like taking out a dead lightbulb and screwing a dimmer one in its place.
There are, I think, at least three schools of thinking when it comes to science fiction prose. Let's be unkind and say that the first school is the Analog approach. This is the notion that the prose, above a certain basic level of competence, has no obligation to be anything other than workmanlike. Cliches, hackneyed turns of phrase, worn-out descriptions, all are sanctioned provided nothing gets in the way of the ideas. The problem, in my view, is that the very dullness of this sort of thing actually works against its intended transparency; it's like a window that hasn't been cleaned. You can sort of see the view through it, but there are lots of cobwebs and smears in the way. In other words, it doesn't do what it thinks it's doing.
I've aspired - and on occasion have no doubt failed - to mostly occupy a middle ground where the prose is aiming for a quality of maximum transparency, a sort of defect-free optical glass. The primary function of the prose is still a delivery system for the story, but it is trying, really trying, to do this with genuine elegance, an economy of expression, some wit and originality, and an avoidance of ugly constructions. CS Forester, for instance, generally wrote this sort of prose, as did HRF Keating. Indeed, throughout the twentieth century it was the default mode for many writers of what we might call the "quality" end of genre writing. It is not trying to do anything really inventive with the language, but at the same time there is an effortlessness to the writing which demands a certain control and authority on behalf of the author. Just as an expert bricklayer can lay a wall with no kinks in it, this is the prose of an expert troweller of words.
The question is, is that the best we can hope for? A few years ago, I was broadly of the opinion that this was not only a good sort of prose for science fiction purposes, but actually the optimum sort. That's not to say that there aren't aesthetic choices to be made at this level - attentive readers of mine, for instance, might note that the prose in Revelation Space generally avoids contractions, something that was important to me at the time. Later, I decided that the avoidance of contractions actually led to an awkwardness which is in itself a bad thing, and so I allowed them back in. Now I am again trying to avoid contractions, but this time at (I hope) a higher level of craft. You live and learn.
Back in 2009, I had the pleasure of sitting on a discussion panel hosted in London by the British Science Fiction Association, in advance of that year's BSFA awards. I can't remember whether we were meant to be discussing the novel or short fiction shortlists, or indeed both, but I do remember that my fellow panelist was Adam Roberts. The first thing to understand about Adam is that he is, and was, astonishingly well-read, to a degree that certainly puts me to shame. I seem to remember, in fact, that Adam stated that he made a point of reading the entire Booker longlist each year. He also keeps up with the major SF longlists. So when Adam makes some remark about the relative merits of science fiction versus the mainstream or literary novel, he absolutely does have the data to back up his statements, and his opinion is worth thinking about.
During that discussion, Adam made a point - I think - that, as good and admirable as lots of SF novels are, as richly as their ideas are explored, it's rare to encounter writing at the same level as the best writing encountered in the modern non-genre novel. The prose, in other words, is often serviceable but it's not doing anything more. The best writing in the non-genre novel is often actively non-transparent; it is quite happy to get between you and the story. When David Mitchell writes of a bat, "chased by its own furry turbulence", he's not shooting for workmanlike. But is that the right mode for SF?
At this point in the discussion Adam and I had a bit of a friendly disagreement. My point, made as well as I could, was that to apply the same set or sets of aesthetic criteria to the SF novel as to the literary novel was in fact a mistake, a category error based on a profound misapprehension of what SF is trying to do. My argument was that it would be equally wrongheaded to apply the aesthetic norms of classical music to, say, punk, because in doing so you would not only misunderstand the terms under which punk operates, but in forcing it to be more like classical music you would rob it of much of its intrinsic vitality.
I think I was wrong, though, and that the category error was mine. Punk is a genre; classical music is unquestionably another. Their boundaries are not especially porous. SF's relationship to literary fiction is more complex than that, more like an embedding or an intersection, and while much of SF does indeed run along genre tramlines, the interesting stuff generally doesn't. The question for me now is to what degree the second kind of prose is still the correct tool for the task, and to what degree I should be pushing beyond it, into what might one call intentional non-transparency.
[Edited - the BSFA event was in 2009, not 2008]